Quantum-Cascade Laser Research

We are trying to develop a new approach in physics and design of nonlinear optical devices based on fully resonant interaction of light with electrons confined in quantum wells and monolithic integration of nonlinear elements and pump sources. Our aim is to combine the tunability and flexibility of nonlinear optical sources with advantages that only semiconductor injection lasers can offer: compactness, high efficiency, injection pumping scheme, low cost, and possibility of monolithic integration with electronic chips. This work is collaboration with groups led by Federico Capasso (Harvard) and Claire Gmachl (Princeton).

A standard setup for nonlinear optical generation includes a passive nonlinear crystal, in which the optical pump beams from one or several high-power external lasers are mixed to generate a nonlinear signal at sum-frequency, difference-frequency, or second and higher harmonics of the fundamental pump frequencies. To avoid absorption, a nonlinear crystal should be transparent for all beams participating in the nonlinear mixing process. This means that all frequencies should be far from any resonances in the nonlinear medium. Non-resonant nonlinear optical susceptibilities are usually quite small. For example, the second-order nonlinear susceptibilities c⁽²⁾ in standard nonlinear crystals are of the order of a few pm/V. Then, to produce an appreciable power, one needs to use long nonlinear crystals and high-power external pumping lasers. The resulting nonlinear optical setup is bulky and expensive (typically $200K).

The lasers that we design and fabricate are injection-pumped quantum-cascade lasers, in which the laser field serves as an optical pump for the resonant, nonlinear optical interaction in the very same active region. In other words, the optical pump source and the nonlinear optical medium are integrated in a single, injection-pumped device. Such a design takes full advantage of giant resonant optical nonlinearities of coupled quantum wells (QWs) that have been observed since late 1980s. Indeed, since the optical pump is generated inside the nonlinear medium, resonant absorption is not a problem anymore. In fact, the pump experiences net gain, not absorption! Therefore, we can have all fields participating in the nonlinear interaction to be amplified and to be resonant with corresponding intersubband transitions. As a result, the second-order nonlinear susceptibility c⁽²⁾ is enhanced by orders of magnitude, up to hundreds nm/V.

Nonlinear processes involved can include sum-frequency and second-harmonic generation, third-harmonic generation, difference frequency generation, Raman lasing, parametric down-conversion, and even lasing without inversion. An example of laser design for second-harmonic generation is illustrated in Figs. 1,2.

(b)

(a)

Fig. 1. (a) One section of QC laser in which active region serves as a resonant cascade for SHG. (b) Active region details. Laser emission is generated between states 3 and 2. Laser frequency is nearly resonant to the transition 3-4. As a result, SH I generated nearly resonant to the transition 2-4. There is another cascade – 3-4-5, which provides a significant contribution to the SH intensity.

Fig. 2. Waveguide design for modal phase matching: TM₀₀ mode at the fundamental frequency propagates with the same phase speed as the TM₀₂ mode at second harmonic. Modal overlap with active region where SHG takes place is maximized. A different method of phase matching based on Stark shift of intersubband electron resonances was demonstrated in [4].

Fig. 3. The spectra of both fundamental laser radiation and the nonlinear signal for two different lasers.

The key figure of merit is an efficiency of transformation of the fundamental pump power into the nonlinear signal power, which is measured as the ratio of the nonlinear power to the fundamental laser power squared. The highest efficiency we achieved so far is 35 mW/W² with the highest nonlinear power of 2 mW [6]. An improvement in both power and efficiency by at least a factor of 10 is feasible for second harmonic generation.

Furthermore, SHG process can be employed in QCLs in order to generate coherent light at very short wavelengths between 1.5-3 mm where QCLs cease to operate. Short mid-infrared range (2-3 mm) contains many important molecular lines whereas near infrared range 1.3-1.7 mm can be used for fiber communications. An advantage of QCLs for telecommunications is that these lasers have unique dynamical and modulation properties stemming from the extremely short life time of the intersubband laser transition – of the order of 1 ps, which is much shorter than the cavity round trip time and the photon lifetime. As a result, QC lasers are not expected to have the relaxation oscillation peak in their small-signal modulation response, and they can be in principle modulated at THz rates. Therefore, the implementation of QC lasers in the near infrared range would provide a unique semiconductor laser source for fiber telecommunications. We have recently proposed and studied such lasers based on InGaAs/AlAsSb heterostructures with extremely deep quantum wells – about 1.6 eV (Cho and Belyanin, 2009); see the figure below.

Raman Injection Laser [8]

In a general scheme of the stimulated Raman scattering sketched in Fig. 4a, the incident photon of energy is converted into a Stokes photon of energy. The underlying physical mechanism is the scattering of the incident fundamental radiation by a certain eigenmode of oscillations in a material, which is self-consistently excited by the parametric interaction between the incident and the scattered light. The wavelength shift between the fundamental and the Raman emission is determined by a resonance frequency of the internal oscillations and can be due to vibrational (phonon), rotational, plasmon or electronic resonances.

(b)

(a)

(c)

Figure 4. (a) A schematic of Raman scattering of laser light (blue arrow) into a lower-frequency Stokes wave (red arrow). (b) Calculated Raman gain spectrum when the detuning D is much larger than linewidth. (c) Stokes gain spectrum calculated for the structure shown in Fig. 5, with the detuning D equal to 15 meV. Blue line – one-photon absorption, pink line – two-photon contribution. .

Stimulated Raman scattering has been observed in all kinds of media: gases, liquids, solids, and plasmas. Until now, a distinctive feature of all Raman amplifiers has been the necessity of an external optical pump. Also, existing solid-state Raman sources universally employ scattering off a vibrational (phonon) resonance in a crystal or fiber. A powerful fundamental laser radiation is usually required to offset a small value of the Raman gain (several 10^-9cm/W).

(c)

(a)

(b)

Figure 5 (a) Active region of a resonant Raman laser that integrates fundamental laser cascade 6-5-4 and resonant L-scheme of transitions 1-3-2-1 for Raman Stokes lasing. (b) Laser (upper curve) and Stokes (lower curve) powers as functions of the injection current. (c) Laser (blue line) and Stokes (red line) spectra.

In recent work [8] in collaboration with Dr. Capasso’s group at Harvard, we demonstrated the first injection-pumped Raman laser, where the fundamental and the Raman radiations are both generated by intersubband electronic transitions in the very same active region of a quantum cascade laser (QCL). The stimulated scattering and lasing are due to the excitation of coherent electronic polarization on the mid-infrared intersubband transition 2-1. Its frequency defines the Raman shift; it has nothing to do with phonons and can vary in a very broad range. It could be also efficiently tuned by a voltage bias since the transition 2-1 is diagonal in real space.

One period of the Raman laser structure reported in Ref. [8] is shown in Fig. 5(a). Laser light at 6.7 mm generated on the transition 6-5 serves as a resonant optical pump for lasing at the Stokes wavelength of 9 mm, which is detuned by 12 meV from the transition 3-2. Resonant absorption of the pump at the transition 1-3 is overcome by amplification in the pump laser section at the transition 6-5. The triply resonant nature of the process makes Raman scattering very efficient: peak fundamental power is only 40 mW at the threshold for Raman lasing (Fig. 5b), which implies very large gain coefficient of the order of 2x10^-5 cm/W per period (6x 10^-4 cm/W for the whole stack of 30 stages) and high efficiency of the nonlinear interaction. The nonlinear conversion efficiency around 30% has been measured.

Room-temperature THz semiconductor laser [9-11]

In collaboration with Dr. Capasso’s group at Harvard, we have designed and demonstrated a room-temperature semiconductor THz source based on resonant intracavity difference frequency generation in a mid-infrared QC laser; see Fig. 1. To improve power extraction, we employed a wedge-shaped waveguide as shown in Fig. 2 and a hyperspherical silicon lens attached to the device facet (Fig. 3). Maximum THz power was about 7 mW. The THz spectrum and power are in agreement with theoretical predictions. To our knowledge, this is the first room-temperature THz semiconductor laser source.

We have also demonstrated a surface-emitting room-temperature THz source in which the terahertz light is coupled out of the waveguide by a second-order grating etched into the laser ridges; see Fig. 4,5. In contrast to sources where the difference-frequency radiation is extracted from the facet, this approach enables extraction of the terahertz emission from the whole length of the device even when the coherence length is small. This geometry should also yield a much better beam quality as compared to edge-emitting devices.

Work is currently underway to further increase THz DFG efficiency and power by improving the active region design, the grating efficiency, and the heterostructure growth quality.

Fig. 4. a) Schematic drawing of the ridge waveguide structure with the gold grating on top. b) The TM₀₀ mode profiles for the THz wave (λ≈60μm) generated via difference-frequency generation (DFG) and the two mid-infrared pumps (λ₁≈8.9μm and λ₂≈10.5μm). The mode profiles were calculated for the waveguide without grating that consist of a 400nm-thick layer of gold, a 200nm-thick InP plasmon layer (n-doped 5x10¹⁸cm^-3), a 3.5μm-thick InP cladding layer (n-doped 5x10¹⁶cm^-3), a 6μm-thick active region, and an InP substrate (n-doped 1x10¹⁷cm^-3). The position of the active region and depth of the grating groves (~300nm) are indicated in red and grey, respectively. The grating is etched through the top plasmon layer and 100nm into the lower doped top InP cladding layer.

Fig. 5. a) The mid-IR emission spectrum collected from the edge of a typical device operated in pulsed mode at 80K. b) The THz emission spectrum from the surface of the device in (a).

Recent publications:

[1] N. Owschimikow, C. Gmachl, A. Belyanin, et al. Phys. Rev. Lett. 90, 043902 (2003).

[2] C. Gmachl, A. Belyanin, D.L. Sivco et al., IEEE J. Quantum Electron. 39(11), 1345 (2003).

[3] O. Malis, A. Belyanin, C. Gmachl, et al., Appl. Phys. Lett. 84, 2721 (2004).

[4] M. Belkin, M. Troccoli, L. Diehl, F. Capasso, A. Belyanin, D.L. Sivco, A.L. Cho, Appl. Phys. Lett., 88, 201108 (2006).

[5] C. Gmachl, N. Owschimikow, A. Belyanin, et al., Appl. Phys. Lett. 84, 2751 (2004).

[6] O. Malis, A. Belyanin, D.L. Sivco, J. Chen, A.M. Sergent, C. Gmachl, and A.Y. Cho, Electron. Lett., 40, 1586 (2004).

[7] T. Mosely, A. Belyanin, C. Gmachl, D. L. Sivco, M. L. Peabody, and A. Y. Cho, Optics Express, 12, 2972 (2004).

[8] M. Troccoli, A. Belyanin, F. Capasso et al., Nature, 433, 845 (2005).

[9] M. Belkin, F. Capasso, A. Belyanin, D. L. Sivco, A. Y. Cho, D. C. Oakley, C. J. Vineis, and G. W. Turner, Nature Photonics, 1, 288 (2007).

[10] M. A. Belkin, F. Capasso, F. Xie, A. Belyanin, M. Fischer, A. Wittman, and J. Faist, Appl. Phys. Lett. 92, 201101 (2008).

[11] C. Pflügl, M. Belkin, Qi Jie Wang, M. Geiser, A. Belyanin, M. Fischer, A. Wittmann, J. Faist, and F. Capasso, Appl. Phys. Lett. 93, 161110 (2008).